Publications / 2020 Proceedings of the 37th ISARC, Kitakyushu, Japan

Ontology-Based Decoding of Risks Encoded in the Prescriptive Requirements in Bridge Design Codes

Fahad Ul Hassan and Tuyen Le
Pages 98-104 (2020 Proceedings of the 37th ISARC, Kitakyushu, Japan, ISBN 978-952-94-3634-7, ISSN 2413-5844)

Bridge designs are typically governed by a voluminous set of requirements in various design standards and codes. The requirements are aimed at ensuring the structural safety against different environmental risks experienced by a bridge facility during its service life. The requirements provided in the bridge design standards are generally prescriptive in nature that do not explicitly specify the types of risks addressed in them. As a result, the understanding of the risks hidden in the requirement text is solely associated with the individual designer who often lacks adequate training in interpreting the risks addressed in prescriptive requirements. The conventional practice of manual identification of risks encoded in the prescriptive provisions requires much effort, domain knowledge and may include human errors as well. Little attention has been paid towards automated identification of risks encoded in the prescriptive requirements. The paper presents an ontology-based risk decoding model to decode the risks implied in the prescriptive requirements. The risks included the earthquake, flood, wind, fire, vessel collision, blast loading, temperature and overloading. An ontology for conceptualizing the domain knowledge of the eight risks was first developed. The ontology-based decoding model ranks the risks for a prescriptive requirement by measuring the semantic similarity between the requirement and the risk ontology. The model was tested on the AASTO bridge design specifications and evaluated in terms of Spearman, Kendall tau and Pearson rank correlation test. This study is expected to assist the designers in the improved understanding of risks encoded in prescriptive design standards.

Keywords: Bridge design; Prescriptive requirements; Design standards; Environmental risks; Natural language processing; Deep learning